There are already a large number of textbooks in econometrics so some reasons are needed to justify another one. This book is based on my experience of teaching econometrics, and being asked for help from various people, with regression output to interpret and models to build, for over 20 years. This has lead me to the conclusion that a new approach to the subject might be useful in dispelling the fear and confusion that surround this subject.
My particular interest in this book is the audience of (sometimes reluctant) undergraduate students for whom econometrics is compulsory and those (usually
Ph.d students in economics) who have taken some econometrics in the past but are finding it very hard to relate this knowledge to their own research work or articles they need to read. The book should also be useful for those who have chosen undergraduate options particularly for forms of assessment based on collecting data and submitting written reports on the results.
I have been inspired by the basic fact that the people mentioned above do not seem to have been able to find what they want in the existing books. This has come as something of a surprise to me as for years I have been recommending books which seemed to me to contain all the relevant knowledge in a suitable format but have increasingly been met with a negative reaction to these books.
The problem seems to be that the existing books do make perfect sense IF you already have a good knowledge of the subject and a fairly solid background in statistics.
This book starts with the assumption that you know nothing about the subject but are at least interested in economics, or other social sciences, and may have some background in basic statistics. The approach I have taken differs in a number of ways from what is usually the case in econometrics texts. I now elaborate on these.
The emphasis on this book is on econometrics in terms of why it exists, and how it is used in daily practice, rather than just as a set body of rules which have to be learnt. This daily practice involves three things- doing econometrics- as in collecting data and running regressions, writing these up into some kind of report and reading such reports written by others. I have attempted to make most chapters of this book an integrated treatment of all three aspects.
I have taken as the starting point the fact that most people’s interest in this subject stems from a need to analyze data that requires them to have tools to do so. Many econometrics texts seem to start from the opposite point of view, which is that econometrics is a branch of statistics, which is of interest purely for its own sake and requires some data as a vehicle to use the tools. Following this principle, the idea of probability and hypothesis testing is introduced in the context of a discussion of types of data rather than as a separate area of formal study.
This text contains relatively little mathematics and also has fewer graphical illustrations than is conventional. Despite this, the book is not intended to be a ‘watering down’ of the subject, as it does address the fundamental issues encountered in applied econometric research. The reduction of graphical material is deliberate as I have found the use of two-dimensional diagrams to illustrate the concepts of econometrics tends to lead to later mistakes in thinking in the context of a multivariate model dimensional data matrix. It would be better, conceptually, if the whole subject was approached in terms of matrices but this does not seem feasible for a general audience. I only provide graphs of actual data-series occasionally as these do not provide that much insight into a regression model. The earlier chapters feature a number of time-series where the pattern is fairly obvious without a graph and the end-of-chapter exercises invite students to draw graphs where appropriate.
The mathematical notation, which is used in this book, has been chosen with the student in mind. I have stuck to the English alphabet in describing parameters –hence we have b and b hat rather than the more usual beta and beta hat. There are various reasons for this. One is that this is a notation the student is likely to be more used to this (in their micro- macro- and maths for economists courses) as a way of writing down models algebraically. A further reason for not using beta is that some software packages present beta coefficients as part of their default output (and some also use the b notation for regression parameters). Using the term beta for regression parameters, in the econometrics texts, leads to students getting confused when they move to computer output.
I have taken the step all through of using a 0 subscript for the intercept and subscripts from 1 to k for the parameters in an OLS regression. This is because numbering up from 1 starting with the intercept would lead the struggling reader to be more confused by finding variable X1 attached to parameter b2 and so on.
I have paid particular attention to the order of the material covered here. All econometrics texts cover pretty much the same basic topics and this one is no exception. However the material comes in a bewildering array of orders, in different books, which far exceeds the diversity found in a core micro- or macro-economics text at the same level. Some books feature subjects at very different locations in the book. I have strived to produce an order, which forms a coherent and stimulating flow for the intended readership. This means that some subjects are given a chapter to themselves when they do not really warrant this in terms of their conceptual importance. The chapter division and section divisions, used are intended to help the continual reinforcement of the learning experience. Thus we have a whole chapter on dummy variables when this is something that could really be disposed of in a few pages if one has a strong grip on the subject. The dummy variables chapter, like most other chapters, is used not just to instruct in the topic of the chapter but also to help build the reader’s confidence in seeing some sort of developing pattern to what they are learning. We thus have some repetition of how to interpret results as we go through the book.
This book adopts a direct conversational style in addressing the reader often dealing with issues in terms of questions that might be asked by the puzzled reader. It also seeks to assist the reader by reminding them which earlier topics may be useful in assisting them on the current topic.
The use of illustrative material is different in this book. I have tried to strike a balance between providing a variety of topics (including sex and drugs and rock and roll) which show how widely econometrics can be applied and avoiding confusion by showing people too many different results.
In my experience students learn econometrics more efficiently when they see techniques being applied to data they have already seen rather than every new topic having an entirely new data set. In teaching econometrics to classes of 100+ (mostly reluctant) students I have resorted to the extreme strategy of using the same data set for the computer class exercises. This does seem to produce remarkable improvements in customer satisfaction amongst other things. It is much less interesting for the instructor and there may be some kind of trade-off between teacher boredom and student comprehension in this particular subject.
Such an extreme strategy, of using the same data all the way through, would not be acceptable in a book but I have taken the same sort of approach by using a limited set of data sets for the illustrative results based on analysis of data, which is available to the reader. For example, non-linear functional forms are illustrated by estimation on the same data that was used earlier to compute partial correlation coefficients, bivariate regression and multiple regressions. The same data set is later used to look at the results of proposed solutions to multicollinearity.
The illustrative results are also used to demonstrate to the novice econometrician that things do not always work out as perfectly as they should and we might even get some very strange looking results. This leads on to an emphasis on how much results do change when you change the specification. This is a very obvious point and one that should be obvious from the existing texts but my experience suggests it is only really grasped by repetition and more likely personal experience of doing regressions. To this end, I have provided end-of-chapter exercises which are integrated with the text. These often ask the reader to attempt different equations on data for which results have already been displayed in the text.
In the text I have given more background on the underlying rationale behind any equation than is normally given as this lack of context often seems to add to confusion. I have largely resisted the temptation to quote results of estimated equations from published literature as people seem to find it hard to digest these shorn of the context in which they were originally presented.
As an alternative, I have developed a template (shown in Appendix 1) suitable for summarizing the essence of a published paper. Most chapters contain examples of how to write such summaries with additional background material on the articles chosen. To get the most out of this, the reader would really need to obtain copies of the source articles. However, this approach is easily configurable to the needs of a particular econometrics module. It would be possible to make the act of filling in the template, using different articles of the instructor’s choosing, a part of the course assessment.
2005 McGraw-Hill Education Europe